Meaning of "nice utilisation" in cpu metrics in glance

[Conal O'Kelly]

Hi
I have noticed in glance that CPU utilisation is almost continuously at 100%. Drilling down in glance "option a", I can see that the Nice state is accounting for about 90% of this utilisation. Can anybody explain what this means?? It is a group of oracle processes that are consuming most of the CPU and all have nice priority of 22. The DBAs have since bounced the database and as a result CPU utilisation is still around 100% but is now showing the User state is accounting for around 90% which is more what I'd expect.
Any insights greatly appreciated.

Cheers
Con

[Bill Hassell]

Nice means lowered priority. It is common to start processes that are compute-intensive and not interactive using the nice command. Higher priority processes will obtain CPU time before the lower priority processes.

The fact that there are a lot of nice processes consuming time is not important. The amount of CPU time consumed by databases is 100% dependent on the efficiency (or lack thereof) of query statements and the number of active clients. There is no need to be concerned if usage stays at 95%, even 100%, as long as users are happy with the results. If the systems starts to seem slow, run a database statistics collection on each of the queries and rewrite the ones that are used the most. Thios also includes adding new indexes and making sure that current indexes are valid and balanced.

[Conal O'Kelly]

Thanks for your response Bill. I thought every process that started had a nice value (generally inherited from its parents process). Does the fact that glance is showing nice utilisation mean that the process has been reniced or started using the nice command or a nice system call?? I'm still unsure of the exact difference between user & nice state for cpu util in glance.

Thanks again for your reply.

Cheers
Con

[Steve Steel]

Hi

If you have 11.11 load this and dependencies.

Patch Name: PHKL_24551

Patch Description: s700_800 11.11 thread nostop, vPar, Psets, load averages

Creation Date: 01/07/18

Post Date: 01/08/02


steve Steel

[Stanimir]

Hi,
Bill very well explain nice priority.
UNIX dynamically calculate priority with next formula:
Priority=(???recent CPU usage???/constant)+ (base priority)+(nice value)
/Maurice J.Bach ???The design of the UNIX operating system???/.
Base priority for HP-UX is 20. All interactive users have only base priority and
nice priority=0.
ORACLE have 4 or 5 base processes: ora_pmon, ora_smon, ora_dbwr ect.
If this processes have nice priority=22 , scheduler planing them after all other.
All users who submit querys to ORACLE must wait these processes ???to end his work???.
If these processes have negative nice priority, scheduler planing them more frequently and
ORACLE work with beter performance.
I use nice priority= -10 for ora_smon, ora_pmon ect.
To improve ORACLE performance you can use HP PRM.